Simple Single-Ended Headphone Amp

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
A piccture of the headphone amp all buttoned up. I'll be auditioning it with my new Grado SR225E headphones this afternoon.
 

Attachments

  • Img_1739X.jpg
    Img_1739X.jpg
    175.9 KB · Views: 474
This amp rocked the Grado headphones with plenty of volume to spare, even from the line output of my Panasonic Discman. The AKG 240 studios were a different matter, requiring me to hook up to the earphone output instead, but those are notoriously difficult cans to drive. I'm satisfied.
 
Member
Joined 2011
Paid Member
Output impedance might be on the high side, making it tougher to drive 16 ohm headphones. If I punched it correctly on my calculator, the impedance of the coupling capacitor C6 alone, is 5.3 ohms at 30 Hz. Add that to the impedance contribution by R10A//R10B (reduced by NFB!!), and the total might become big.
 
I strongly suspect that the output impedance of this amp is pretty damn low, even though there are only 4 active devices in the game. The current source loading on the input stage, the lack of significant degeneration in the output mosfet, and the overall high bias current levels result in a high amount of open loop gain, fed back to produce a small closed loop gain. The closed loop gain, though on the low side, is sufficient to drive a pair of 30 ohm Grado headphones to earsplitting levels using the line output of my Panasonic Discman. The low THD levels in my simulations for 50 ohm load also indicate a low output impedance/high open loop gain.

When I get back to work (where my decent copy of PSpice resides), I'll try tinkering with the output loading to determine the effect on output amplitude and distortion. I suspect that if I increase the output coupling cap appropriately, a 16 ohm load might not trouble it at all. BTW, what audiophile headphone out there has that low of an impedance?

During my vacation, I have busied myself with ginning up a layout that uses currently available parts. I used the MPSW45A as input device in the proto that I'm shipping, as it is in the TO-92L package that can dissipate about 0.5W without significant sweat, and the current layout wants a TO-92 footprint. Unfortunately, all the nice TO-92L package devices have been obsoleted, so I've been forced to sub a SOT-223 device in the new layout for both the input device and the mosfet providing low impedance voltage drop in the source of the output mosfet. The layout was also tweaked to accommodate a 2200uF output coupling cap that should be suitable for 16 ohms load, and also removes one line item from the BOM.
 
OK - I did some searching around, and the 16 ohm impedance seems to be common among phones meant for jacking into I-whatchamacallits and the like. On Monday, I'll see what effect the lower impedance has on THD given an appropriate output coupling cap. I'll lower the exciting voltage to compensate for the lower output impedance, as the rationale for the low impedance is the low output voltage available from mobile devices.
 

PRR

Member
Joined 2003
Paid Member
I found the tables on this web page to be incredibly helpful when planning out the design of a headphone amplifier...
Headphones sensitivity, impedance, required V/I/P

Long ago, I compiled data from RANE for an unhealthy 120dB goal (Real Frikkin LOUD!!) and plotted on V/I axis, with Z and P as slant-lines.

For a less-insane 110dB SPL, all V and I are 1/3rd, all P is 1/10th of plot.

While only a few of the models dotted are still on the market, the technology is mature and present models will be in this flock, though some trend to the 32 Ohm cluster for iPod use.

If you are aiming an amplifier to cover "any headphone", this gives some guide to how much Voltage supply and current delivery you must provide.

For less-intense 110dB SPL, ignoring some outliers, 4V 60mA (RMS) covers "most" headphones. Those outliers are mostly expensive and notoriously inefficient, so not a first choice if max SPL is an issue. (I could live the rest of my life never turned-up over 90dB SPL.)
 

Attachments

  • I-V-Z-P.gif
    I-V-Z-P.gif
    17.4 KB · Views: 319
I did a sim using 16 ohm load for the original circuit (with output cap increased to 2200uF), and though the output amplitude was the same as for 50 ohms, the THD was up to 0.1%, though predominantly 2nd harmonic. Tweaking some resistor values to get the quiescent current higher improved matters quite a bit. This may be the option I pursue when I get my 2nd rev boards in house.
 

Attachments

  • SE_HP_16R0_SCH.GIF
    SE_HP_16R0_SCH.GIF
    20.8 KB · Views: 433
  • SE_HP_16R0_DIS.GIF
    SE_HP_16R0_DIS.GIF
    13.8 KB · Views: 435
Last edited:
Debug finished - me and Shrut enjoyed the sound with various jazz inputs mediated through a Panasonic Discman and 30 ohm Grado headphones. I think I will split the difference between this version and the former to settle on an an output bias current of ~ 150 mA. This will be easier on the output bias fet and gain setting resistors. I don't feel like catering to I-zombies with this simple amp - they can use a more ambitious amp for driving their 16 ohm whatchamacallits endorsed by people who plug music not requiring Hi-Fi drivers. More later.
 
Here's what I'm considering the final version of this amp. The bias current in the output fet is~170 mA, and everything should be far less toasty.
Included is a sim with bias points and simulated distortion for loads of 16 ohms, 30 ohms, and 50 ohms. I'll bring the amp into work soon to get the gain-phase plots. I also plan to drive it with a more ambitious source than my little Panasonic Discman to see if the source material shines through.

After that, I'll concentrate on four other options I have in mind, some sand state, others vacuum.
 

Attachments

  • SE_HP_51FB_SCH.GIF
    SE_HP_51FB_SCH.GIF
    19.7 KB · Views: 357
  • se_hp_dis_51_16ld.GIF
    se_hp_dis_51_16ld.GIF
    9.7 KB · Views: 349
  • se_hp_dis_51_30ld.GIF
    se_hp_dis_51_30ld.GIF
    9.7 KB · Views: 319
  • se_hp_dis_51_50ld.GIF
    se_hp_dis_51_50ld.GIF
    9.2 KB · Views: 97
Here are the gain-phase plots for the latest incarnation of this amplifier. I could probably get a bit better phase margin if I started my gain roll-off earlier, but I'm very tempted at this point to declare victory and go home. Zero dB crossover is almost exactly 5MHz for both channels with phase margin about 60 degrees. I'll post some square wave response pictures when I get the time. The gain-phase plots were dome with 0.1V excitation and a 51 ohm load.
 

Attachments

  • SE_HP_51ohm_CH1_G-P.jpg
    SE_HP_51ohm_CH1_G-P.jpg
    178.4 KB · Views: 86
  • SE_HP_51ohm_CH2_G-P.jpg
    SE_HP_51ohm_CH2_G-P.jpg
    185.2 KB · Views: 78
Member
Joined 2011
Paid Member
Looks like a closed loop bandwidth of 0.6 MHz and a closed loop gain of 6X (15.6 dB). I imagine its square wave response (10%-to-90% risetime) will be a wee teeny ooch less than 600 nanoseconds, risetime & falltime.

If the amp circuit is well-approximated by a classical two pole "second order system" then the plot from November (link to post) ought to give a pretty decent prediction of phase margin. Just measure overshoot% when driving the input with a FAST (much less than 600 nsec rise/fall time), low amplitude square wave, then simply read the graph. No need to open the feedback loop.

Why low amplitude? To stay away from the amplifier's slew rate limits.

_
 

Attachments

  • overshoot.png
    overshoot.png
    17.3 KB · Views: 80
Last edited:
Here are the rise and fall times for the current incarnation of the headphone amp - not too shabby for a resistor loaded Class A circuit with only four active parts (and one is just there to drop some voltage).
 

Attachments

  • sehp-51ohm-trise.png
    sehp-51ohm-trise.png
    48.3 KB · Views: 87
  • sehp-51ohm-tfall.png
    sehp-51ohm-tfall.png
    47.2 KB · Views: 79
Member
Joined 2011
Paid Member
not too shabby for a resistor loaded Class A circuit with only four active parts

You could replace M8 with a 1N5334B zener diode (3.6V, 5W) and claim only three transistors. And you could also replace M2 by an E153 Constant Current Diode @ 15 mA (sold at Mouser, part no. 954-E-153) and claim only two transistors. That would get people talking!

Datasheets attached.

_
 

Attachments

  • 1N5334B.PDF
    130 KB · Views: 48
  • E153_datasheet.pdf
    600.4 KB · Views: 53
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.